TY - JOUR
T1 - Deep-learning Object Detection for Resource Recycling
AU - Lai, Yeong Lin
AU - Lai, Yeong Kang
AU - Shih, Syuan Yu
AU - Zheng, Chun Yi
AU - Chuang, Ting Hsueh
N1 - Funding Information:
This work was supported in part by the Ministry of Science and Technology of Taiwan, R.O.C. under Contracts MOST 108-2622-E-018-001-CC3, MOST 108-2221-E-018-017, and MOST 108-2218-E-005-010.
PY - 2020/7/17
Y1 - 2020/7/17
N2 - Recent years have seen a growing concern over global warming, as well as environmental pollution and protection issues. Resource recycling helps the effective reduction of greenhouse gases and environmental pollution, and improves the quality of life for many people. This paper proposes a deep-learning object detection system for resource recycling. The resource recycling of the objects including paper cups, plastic bottles, and aluminum cans was conducted by artificial intelligence. Single shot multibox detector (SSD) and faster region-based convolutional neural network (Faster R-CNN) models were utilized for the training of the deep-learning object detection. With regard to data set images and training time, the accuracy, training steps, and loss function of the SSD and Faster R-CNN models were studied. The accuracy and loss characteristics of the deep-learning object detection system for resource recycling were demonstrated. The system exhibits good potential for the applications of resource recycling and environmental protection.
AB - Recent years have seen a growing concern over global warming, as well as environmental pollution and protection issues. Resource recycling helps the effective reduction of greenhouse gases and environmental pollution, and improves the quality of life for many people. This paper proposes a deep-learning object detection system for resource recycling. The resource recycling of the objects including paper cups, plastic bottles, and aluminum cans was conducted by artificial intelligence. Single shot multibox detector (SSD) and faster region-based convolutional neural network (Faster R-CNN) models were utilized for the training of the deep-learning object detection. With regard to data set images and training time, the accuracy, training steps, and loss function of the SSD and Faster R-CNN models were studied. The accuracy and loss characteristics of the deep-learning object detection system for resource recycling were demonstrated. The system exhibits good potential for the applications of resource recycling and environmental protection.
UR - http://www.scopus.com/inward/record.url?scp=85089474849&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85089474849&partnerID=8YFLogxK
U2 - 10.1088/1742-6596/1583/1/012011
DO - 10.1088/1742-6596/1583/1/012011
M3 - Conference article
AN - SCOPUS:85089474849
VL - 1583
JO - Journal of Physics: Conference Series
JF - Journal of Physics: Conference Series
SN - 1742-6588
IS - 1
M1 - 012011
T2 - 2020 5th International Conference on Precision Machinery and Manufacturing Technology, ICPMMT 2020
Y2 - 3 February 2020 through 7 February 2020
ER -